Particle Swarm Optimization Algorithm And Its Application In Wind Wind Resources Assessment

It's an important measure to develop one of the most promising clean energies under the great background of people worldwide emphasizing the low carbon economy, energy saving and emission reduction. As we know, wind resource evaluation is the most basic work to explore wind power. So, it's a great crucial work to how to accurately and effectively assess wind resources in a region. Wind speed probability distribution is taken as the very important indicator to wind resources, therefore finding out the assessment of wind speed probability distribution can make us know the wind resources status in this region. For the defects with poorer while using the the least square method, moment estimation, maximum likelihood, etc to assess the wind speed probability distribution, it's quite essential to explore a more effective method in order to meet the high precision requirement of the actual wind field. So, this paper tries to propose a new method to fit the actual wind speed curve from intelligence point, such as the improved particle swarm optimization algorithm.As a new intelligent optimization algorithm, particle swarm optimization (PSO) is also a probability-based heuristic search algorithm, and its prime idea comes from social information sharing and information evolution between individuals from the group. In this paper, particle swarm optimization algorithm is researched deeply. First of all, the effects of the parameters of PSO to algorithm performance are studied quantitatively from algorithm principle. Then, the convergence of PSO track conditions are discussed in detail and experimental verification on convergence region is given. What's more, for the single type and complication of evolution model of the traditional PSO, a simplified classification PSO algorithm is proposed, and the optimization performance of this improved algorithm is demonstrated by the simulation results of benchmark testing functions. The main researching results are as following:①. Firstly, a systematic summary is given, such as the research status, application status and development trend of PSO. Then, the basic PSO algorithm and two standard particle swarm optimization are reviewed explicitly and the effects of the parameters of PSO to algorithm performance is analyzed. What's more, the convergence of PSO track conditions are discussed deeply, and the convergence of PSO track conditions is given. In the end, under the one-dimensional solution space, particle trajectories trend changing figures of single particle are finished based on five groups of different parameters according to boundary conditions of parameters convergence, and the correct performance of boundary convergence conditions are verifed by the simulation results.②. A simplified classification of population particle swarm optimization is proposed, basing on the defects of global optimal model falling into local optimum easily and local model with slow convergence speed during traditional PSO algorithm solving a complex multi-dimensional, the peak function. At fitst, particles are classified three types such as the better, ordinary and worse according their size of fitnesses, and then, the velocity item of basic PSO equations is reasonably analyzed and removed, and the evolution equation of PSO is simplified. Based the work above, the particles of three types are adjust dynamically using three types corresponding, such as, cognitive model, complete model and social model. In these three models, so the demand difference of different particles is balanced effectively on the local search ability and global search ability. Through a typical simulation experiment of four test functions and comparison with the classical PSO and an more satisfactory improved PSO at present, rationality and effectiveness is showed according to the examples.③. Traditional statistical methods, such as, least squares method, moment estimation, maximum likelihood, are analyzed in order to make a model and optimize on these two parameters, when wind speed is considered to obey two-parameter Weibull distribution. A conclusion can be obtained that traditional statistical methods are easily achieved but with poor accuracy. So it's essential to explore a new effective computation method, and a new method is firstly proposed based on a intelligence point, such as using modified particle swarm optimization algorithm to make an model and optimization over the two parametes of Weibull distribution, including combination with practical field wind power datas. Compared to the results figured out by conventional least-squares method, Denmark WAsP software and historical wind speed data sequences for the computation of wind indicators reflecting the wind energy resource characteristics, higher precision is shown in this suggested method according to the example in this paper, and closer to actual wind condition and can have more practical reference.